The most common stream of tokens where every token is buffered up and tokens are filtered for a certain channel (the parser will only see these tokens). Even though it buffers all of the tokens, this token stream pulls tokens from the tokens source on demand. In other words, until you ask for a token using consume(), LT(), etc. the stream does not pull from the lexer. The only difference between this stream and BufferedTokenStream superclass is that this stream knows how to ignore off channel tokens. There may be a performance advantage to using the superclass if you don't pass whitespace and comments etc. to the parser on a hidden channel (i.e., you set $channel instead of calling skip() in lexer rules.)